IL275859B1

IL275859B1 - Method and device for improving bandwidth utilization in a communication network

Info

Publication number: IL275859B1
Application number: IL275859A
Authority: IL
Inventors: Krayden Amir; Sadeh Or; Zakin Ori; Freilikhman Gregory
Original assignee: Drivenets Ltd; At& T Services Inc; Krayden Amir; Sadeh Or; Zakin Ori; Freilikhman Gregory
Priority date: 2018-01-28
Filing date: 2019-01-19
Publication date: 2024-01-01
Also published as: EP3744066A1; WO2019145938A1; US20200351222A1; IL275859A; EP3744066A4; US11444890B2

Description

METHOD AND DEVICE FOR IMPROVING BANDWIDTH UTILIZATION IN A COMMUNICATION NETWORK TECHNICAL FIELD The present disclosure relates generally to the field of communication network throughput, and in particular, to a device and a method for alleviating bottlenecks in communication networks, thereby enabling achievement of an increase in the communication systems' throughput. BACKGROUND Every computer included in a data network, operating either as a client or as a server, requires a network interface card (hereinafter "NIC") in order to access the network. A NIC is usually a separate adapter card that is mounted onto one of the server’s motherboard expansion slots. However, most newer computers have the NIC built into the motherboard, in which case a separate card is not required. For a client computer, one may use an inexpensive built-in NIC, because a client computer is used to connect only one user to the network, contrary to the use of a NIC in a server computer which connects many network users to the server. Smart NICs may be used in cloud data center servers to boost performance by offloading CPUs in servers by performing network datapath processing. With the recent shift in cloud data center networking driven by software-defined networking (SDN) and network functions virtualization (NFV), a new class of offload NIC is needed - namely a smart NIC. Specifically, there are three major reasons for which a smart offload NIC, or a smart NIC, would be required: 30 1) The complexity of the server-based networking data plane has increased dramatically with the introduction of overlay tunneling protocols such as VXLAN, and virtual switching with complex actions. 2) Increasing network interface bandwidths means that performing these functions in software creates an untenable load on the CPU resources, leaving little or no CPU left over to run applications. 3) A key requirement of SDN is that the networking data plane must remain fungible, so fixed-function offload technologies cannot be applied. Data networking involves both hardware and software. On the software side, networking protocols are often designed for current (or near-term) hardware capabilities. Protocols often become widely adapted while at the same time networking hardware improves. Processors become more efficient while communication mediums gain reliability and increased capacity. However, over time, networking protocols may become less suitable for the hardware that is being developed. Moreover, some capabilities of networking hardware have not been fully appreciated and realized. Smart NICs, for example FPGA (Field Programmable Gate Array) NICs, have become more common. These newer interfaces, like traditional NICs, provide physical and media connectivity. They also include additional processing capability, sometimes in the form of reconfigurable circuitry (e.g., FPGAs). These processing-augmented NICs may allow features of some protocols to be offloaded from the host's CPU (Central Processing Unit) to the NIC. Some smart NICs may even allow an entire transport protocol to be fully offloaded from a host's CPU to the smart NIC. However, this approach often requires significant host-side changes. For example, host-side software might need to be re-written in order to enable communicating directly with its NIC via a custom application programming interface (API) rather than via a standard transport protocol. US 20160381189 describes a smart NIC with features that enable the smart NIC to operate as an in-line NIC between a host's NIC and a network. The smart NIC provides pass-through transmission of network flows for the host so that packets sent to and from the host would pass through the smart NIC. As a pass-through point, the smart NIC is able to accelerate performance of the pass-through network flows by analyzing packets, inserting packets, dropping packets, inserting or recognizing congestion information, and the like. However, the main bottleneck in network applications is generally the I/O bandwidth. In a case wherein NICs are installed in servers, it is the PCI (Peripheral Component Interconnect) bandwidth that limits the maximal throughput of the system. A conventional PCI is a local computer bus used for connecting hardware devices in a computer. PCI is part of the PCI Local Bus standard. The PCI bus supports the functions found on a processor bus but in a standardized format that is independent of any particular processor's native bus. The present disclosure seeks to solve the drawbacks described above.

Claims

1.WHAT IS CLAIMED IS: 1. A communication system comprising at least one smart network interface card ("NIC") provided with a logic/programmable processor and a local memory, and a computing element, wherein a communication bus is used to connect said smart NIC and said computing element to enable forwarding data there-between, wherein said system is characterized in that said smart NIC is configured to receive a data packet, to extract metadata therefrom, wherein said metadata resides within said data packet header and/or at one or more specific locations within the data packet payload, place the extracted metadata into a descriptors MetaPacket structure, and forward the descriptors MetaPacket structure, to said computing element along said communication bus, thereby forwarding less than all of said data packet to said computing element; and wherein said system is further characterized in that the computing element is configured to process said descriptors MetaPacket structure and to return the processed descriptors MetaPacket structure to the smart NIC.

2. The system of claim 1, wherein upon receiving the processed extracted metadata, said smart NIC is further configured to extract a respective data packet that had been stored at the local memory and re-write a header of the retrieved data packet according to the processed metadata.

3. The system of claim 1, wherein a plurality of the received data packets is stored at said local memory of the smart NIC.

4. The system of claim 1, wherein the dedicated descriptors comprise one or more members of the group that consists of: (1) the memory location of where the packet was stored, (2) timestamps, (3) statistics, and (4) packet validity flags.

5. The system of claim 1, wherein the smart NIC processor is further configured to receive a plurality of data packets and to identify one or more data packets from among the plurality of data packets received, for which an entire content is required, and to forward the entire content of said identified data packets towards the computing element.

6. The system of claim 1, wherein the smart NIC processor is further configured to apply an aging mechanism to packets that are stored at the local memory of said smart NIC.

7. The system of claim 1, wherein upon processing a descriptors MetaPacket structure by a software application residing at the computing element, the processed descriptors MetaPacket structure is forwarded along a reverse path of said communication bus, to be used by the smart NIC processor in extracting a corresponding data packet from the local memory and processing the extracted corresponding data packet according to the descriptors comprised in the processed descriptors MetaPacket structure.

8. The system of claim 1, wherein upon determining in accordance with updated information contained in the descriptors MetaPacket structure that the data packet should be forwarded from the smart NIC through an egress port which is other than an ingress port via which that data packet was received, forwarding the data packet to a driver which in turn forwards the data packet to an appropriate egress port.

9. The system of claim 1, further comprising a driver configured to support two modes of operation, a transparent mode and an explicit mode.

10. A communication system comprising at least one smart network interface card (“NIC”) provided with a logic/programmable processor and a local memory, and a computing element, wherein a communication bus is used to connect said smart NIC and said computing element to enable forwarding data there-between, wherein said system is characterized in that said smart NIC is configured to receive data packets, to extract data therefrom and to forward less than all data comprised in the received data packets, to said computer element along said communication bus, wherein said communication system further comprises a driver configured to support two modes of operation, a transparent mode and an explicit mode, and wherein the driver is configured to reconstruct MetaPackets by using metadata associated therewith, while leaving the remaining of the packet nullified.

11. The system of claim 10, wherein said driver is further configured to forward MetaPackets to a smart NIC, which in turn is configured to operate on MetaPackets received from the driver.

12. A method for use in a communication system that comprises at least one smart network interface card ("NIC") provided with a logic/programmable processor and a local memory, and a computing element, and wherein a communication bus is used to connect said smart NIC and said computing element to enable forwarding of data there-between, said method comprises the steps of: (i) extracting metadata from a data packet that arrives at the smart NIC and placing the extracted metadata into a descriptors MetaPacket structure; (ii) forwarding the descriptors MetaPacket structure to a software application CPU; (iii) processing the descriptors MetaPacket structure by the software application CPU and returning the processed descriptors MetaPacket structure to the smart NIC; (iv) retrieving a respective data packet that has been stored from the local memory; (v) re-writing the data packet's header according to updated metadata; and (vi) forwarding the data packet via an egress port.

13. The method of claim 12, wherein said descriptors MetaPacket structure comprises at least one of: a data packet's header and an address within the local memory of the stored data packet. For the Applicant, By: